by Paul Braterman
Evolution has nothing to do with progress. Most evolution doesn't even have anything to do with adaptation, and it is perfectly possible for a change that is worse than useless to spread through a population. Paradoxically, however, such non-adaptive change may be a necessary prelude for major adaptations.
This post was inspired by a recent opinion piece (open access here1) in BMC Biology, entitled "Splendor and misery of adaptation, or the importance of neutral null for understanding evolution" (I will explain what "neutral null" means later). The paper itself is in parts highly technical, with 86 references to the original scientific literature, but I will try here to give a general overview of some of the main conclusions, and to place them in context.
Darwin and Wallace both thought that evolution was driven by selection. If so, then whenever we find a feature in an organism, it makes sense to ask what function it serves. The function may for example be help in survival (natural selection, in the narrowest sense of the term), or help in obtaining mating opportunities (sexual selection).
Because the evolution of a species is constrained by its history, there will be features that are themselves non-adaptive, but come about as side-effects of more important adaptive changes. Such incidental maladaptions include the tortuous paths of major nerves and arteries, which have arisen as the unwanted by-product of changes in body plan since our fish-like ancestors. One well-known example is the recurrent laryngeal nerve, which loops under the aorta near the heart and back up again on its way from the cranium to the larynx and oesophagus. In a fish, its path is more or less a straight line, but as the heart has moved down in the body, and the aorta with it, the nerve has been forced into this contorted pathway.
Likewise, we can expect to find vestigial organs, which once had a function, and are now redundant, but have not yet completely disappeared. An example is the pelvis of the whale, inherited from its four-legged ancestors. Such vestigial organs often acquire secondary functions, in the phenomenon known as exaptation. The bones in the mammalian ear, related to bones in a reptile's flexible jaw, illustrate this. [Insert diagrams: whale pelvis; jaw-to-ear] And indeed whales use their pelvis and femur relics in sexual embraces.
Adaptationism is the view that all aspects of an organism are, directly or indirectly, the result of selection. So every feature needs to be explained, either in terms of its own function, or as an incidental relic or side-effect of more directly functional features. This is a natural enough assumption, but like all assumptions it requires justification. Otherwise it is merely a "Just So" story.
Before committing ourselves to finding a function in any specific case, we need to show good reason 
L: How the elephant got its trunk, from Kipling's Just So Stories, 1902
It is now widely accepted that features can arise from no particular reason, as a result of what is called genetic drift, the term for accumulated random change in the genome that confers no evolutionary advantage. This idea emerges naturally from the discipline of population genetics, which for a century now has been modelling, and studying by observation, the arithmetic of how mutations spread or diappear in populations. Many readers will be familiar with Theodosius Dobzhansky's 1973 observation, "Nothing in biology makes sense except in the light of evolution", but to this we must now add the rider put forward by Michael Lynch in 2007, that "Nothing in evolution makes sense except in the light of population genetics." This is part of a major change in perspective. Evolutionary change is the result of drift, as well as of selection. This means that we should stop thinking of evolution as a process by which populations become more fitted to niches, but as something more like random diffusion among possibilities. Which genetic variations will die out, and which will spread through the population, becomes a matter of statistics.
Nothing in biology makes sense except in the light of evolution – Dobzhanski, American Biology Teacher, 1973
Nothing in evolution makes sense except in the light of population genetics – Lynch, PNAS, 2007
Now consider the chance of something going wrong. New variations arise all the time through mutations, the errors that occasionally occur when DNA is copied from one generation to the next. The chance of a particular mutation happening is completely independent of whether it is beneficial or harmful. And harmful mutations will happen much more often than those that are beneficial, simply because there are far more ways of making things worse, than there are of making things better. For example, if you change a word at random in a sentence, you are unlikely to improve the meaning.
So what can we say about the chances of a harmful mutation spreading through a population? It turns out that the odds here depend on two things. One, obviously, is how harmful the mutation is, in terms of reproductive success, which is the only thing that counts here. The more harmful the mutation, the smaller the chance of its taking hold. The other one is population size, and the effect here is not quite what I would have guessed.
My naive expectation was that selection pressures would be faster and more effective with smaller populations, simply because there would be fewer inferior individuals to outcompete. I was only half right. Genetic change can certainly take place more rapidly in smaller populations, and the punctuated equilibrium theory of Gould and Eldredge lays heavy emphasis on this fact. But it is far from certain that the change will be beneficial. A smaller population is at greater risk of accidentally committing itself to a damaging change. (This, incidentally, is the reason why endangered species are so difficult to rescue, once the population has fallen below a certain size.) Larger populations are better protected from maladaptive drift, because the less favourable variant would need to beat the odds more often in order to accomplish its takeover. The arithmetic here is rather like the arithmetic of gambling; in the short run, some players at the roulette table are ahead of the game, but the casino always wins in the end.
This has surprising implications. For bacteria,2 with population sizes in billions, evolution is very effective. For animals, an interbreeding population will typically number at most some tens of thousands, rendering them far more vulnerable to accidental genetic deterioration, and less certain to incorporate an improvement when it arises.
That sounds ridiculous. Bacteria have remained bacteria, single celled entities, very good at doing what the particular strain has evolved to do (which can range, depending on the strain, from metabolising minerals to eating human flesh), but without the spectacular variety of form and lifestyle that multicellular eukaryotes like us find so interesting. So how did the complexity of the animal kingdom ever come into being, if animal evolution is, relatively speaking, inefficient?
To answer this question, we need to look more closely at the process of evolution. The range of evolutionary possibilities is often compared to a landscape, with peaks of fitness separated by valleys. Selection will push a population upward along a slope, until it finds itself at a peak, when any further change will be selected against. At this stage, evolution becomes a strongly conservative force. As long as the bacteria live under stable conditions, the bulk of the population will be closely confined to the area round the fitness peak. Change the conditions, however, and things can become very favourable for mutations that enable a bacterium to cope with the change, as we know to our cost in the rapid evolution of resistance to antibiotics. We can see this process at work in the Harvard "evolution on a plate" demonstration. The spread of bacteria across an agar-coated plate is halted when they come up against dissolved antibiotic, until a resistant individual emerges by mutation, and continues the forward advance; see here.
Now switch focus, and regard the genome itself as subject to evolutionary pressures, as well as being the carrier of the information on which evolution operates. Imagine a change that lengthens the DNA, without affecting the amount of useable information that it transmits. This can and does happen in many ways. A stretch of DNA might be copied twice in going from one generation to another, or DNA that originated with a virus might become incorporated into the genome and just sit there without contributing to function. These lengthenings will come at a cost, albeit not a very large one; when it reproduces, the organism with a longer genome will have to spend slightly more resources to copy its DNA, but with no attendant benefit. Unused bits of genome material get bred out under the strict evolutionary regime governing bacteria, but can survive for tens of millions of years in animals, and be used to trace family relationships. They confirm, for example, that humans are more closely related to gorillas than to gibbons, and that whales are related to hoofed land animals.
The overall effect of insertions and duplications has been to make the genomes of plants and animals much more complicated than they need to be. Current estimates are that less than 2% of the human genome is directly useful, in the sense of coding for proteins, while another 10% or so has regulatory function, in deciding such important questions as when more directly useful genes are switched on and off, and the rest is junk.3
But junk is not garbage. Garbage is useless, but junk is stuff for which there is no use at the moment. And the junk itself is mutating, more rapidly in fact than material currently in use, because there are no selection pressures to remove errors. So from time to time a sequence that started off as useless may acquire a function, and this function can itself in time become essential. There are even bits of virus-derived DNA playing essential roles in the development of the mammalian placenta.
The evolutionary flexibility offered by junk DNA comes at a cost. It leaves animals with a problem not shared by bacteria, namely how to make sure that the right bits of the genome, and only the right bits, are activated. Animals4 have evolved elaborate apparatus to deal with this problem, as discussed in more detail in the paper that inspired this post.
Surplus DNA derived from duplications has played a dramatic role in evolution. Thus the photosynthetic apparatus of plants contains two major components, imaginatively entitled Photosystem I and Photosystem II, each with different roles to play in the conversion of lightto useful chemical energy, and investigation of DNA sequences shows that these systems evolved from twin copies of a single ancestor.
Or consider the three colour vision enjoyed by Old World monkeys and their descendants, including us humans. Most other mammals can distinguish blue from green, but are red-green colourblind. Now take the DNA that specifies the light-harvesting molecule that makes one kind of receptor sensitive to green, and duplicate it. The spare copy is free to mutate, and there is one single mutation that will shift the sensitivity of its product from green to red. This is exactly what seems to have happened.5
The most extreme possible form of duplication is duplication of the entire genome. This happened, not once but twice, around 450 million years ago, some time after vertebrates had separated from sea squirts, but before the first vertebrates emerged on land. Most of the duplicates have disappeared, but clear evidence of the four-fold replication survives in the order in which genes are strung out along the genomes of modern vertebrates. And all surviving vertebrates, including us, owe their existence to this one creative error.
1] Open Access (Creative Commons 4.0), may be freely copied but must be credited. All links in this post are open access except when stated otherwise.
2] And prokaryotes in general
3] A project called ENCODE, which demonstrated some degree of biochemical activity for 80 percent of the human genome, has been widely misunderstood as refuting the concept of junk DNA by showing that all this 80 percent is recognised by other molecules. In fact, most of this activity is irrelevant.
Conclusive evidence against the ENCODE claim comes from the relative sizes of genomes, which bears no direct relationship to the size or complexity of the organism. For instance, T. Ryan Gregory's onion test: the onion genome is five times as long as the human, but onions are not five times as complex.
4] And eukaryotes in general
5] Paywall. Abstract open access but uninformative.